Incorrect inferred type when indexing Pandas dataframe with column list

[janosh]

Summary

import pandas as pd

df = pd.DataFrame({'A': [1, 2, 3], 'B': [4, 5, 6], 'C': [7, 8, 9]})
cols = ['A', 'B']
sub_df = df[cols] # inferred type is Series[Any] but should be DataFrame[Any] or maybe even DataFrame[int]

$ uv pip show pandas
Using Python 3.13.0 environment at: /Users/janosh/.venv/py313
Name: pandas
Version: 2.3.1

maybe slightly related #851

Version

ty 0.0.1-alpha.19 (e9cb838 2025-08-19)

[AlexWaygood]

Thanks for the report! Yes, this is exactly the same as #851 -- we don't support type alias yet, and pandas-stubs uses type aliases for its DataFrame.__getitem__ overloads; these two things together mean we end up picking the wrong overload

[carljm]

Confirmed this will be fixed by astral-sh/ruff#21394